Soil quality considers the nature, properties or attributes of the soil at a determined place in relation to its functionality for a certain purpose. Those attributes or properties are the result of the interaction of different biotic, physic and human induced factors; nevertheless, the study of all possible factors, which determine soil quality, is not possible. Some factors are more important than others for a certain soil use, and how to select such factors and their relative weight in contributing to define soil quality for a certain purpose is generally difficult to define. Also, soil attributes are correlated to each other to a certain degree, which makes conventional statistical methods hard to be applied. This study proposes the combined use of two multivariate statistical tests, Principal Components (PCA) and Cluster Analysis (CA) for optimal variant selection and soil quality for vegetation cover classification. Texcoco lake in Mexico has been gradually dried out since the time of the Aztecs (≈1400). At the early 70's, a large area (70 km2) of bare soil with little vegetation was the only thing that remained of such lake. For our sampling site, 56 km2 of such area were selected. Salt emergences from a shallow saline aquifer have caused an extreme affectation of soils by salinity. The variation on surface (1-15 cm) soil salinity at our sampling site is very large, from extremely large electric conductivities (ECe) (1 319 dS*m-1), due to the presence of saline crusts, to moderate ECe (11.7 dS*m-1). Salt emergences form crusts of different colors and patterns. Saline grassland is the predominant type of vegetation in the study area. Soil horizon description and samples were taken at 86 sites with different salt and sodium affectation. Soil properties like pH, %sand, %silt %clay, organic matter, CaCO3 contents, electric conductivity in the saturation extract (ECe) and Sodium Absortion Ratio (SAR) were estimated for each horizon. Also other variants like depth to tixotropic horizon, physiological depth and plant available water contents were estimated. Horizon weighted averages of soil variables were estimated for two soil depths: 100 cm and 40 cm. A matrix of average soil characteristics was used in a PCA to identify the role played by all variants within the maximum possible explained variance for both depths. Site scores from the PCA for both depths were then used in the CA to obtain soil classification. Class confidence intervals for all variants were estimated and vegetation associations were used to set soil quality class characteristics. Variance explained by 5 components was higher for depth of 40 cm (95.2%) than the one explained by the analysis considering the whole profile (100 cm). Variants like EC, SAR, tixotropic horizon depth, % sand and % clay were the most important factors defining soil class characteristics. CA dendrograms showed 14 soil quality classes identified for the 100 cm depth and 12 classes for the 40 cm depth. By analyzing class confidence intervals of all variants and the presence of certain species associations (with known salinity tolerances), it was clear that class separation considering only the first 40 cm and a selected group of variants accounted for 95.2% of the total variance and gave the best possible class separation. Soil quality variant selection, and class separation using the combined multivariate analysis proved to be a useful simple procedure to classify soil quality in salt affected sites.